The four known diiodine complexes have distinct geometries. These turn out, as we demonstrate by a bonding analysis, to be a direct consequence of diiodine acting as an acceptor in one set, the van Koten complexes, and as a donor in the Cotton, Dikarev, and Petrukhina extended structure. The primary analytical tool utilized is perturbation theory within the natural bond orbital (NBO) framework, supported by an energy decomposition analysis. The study begins by delineating the difference between canonical molecular orbitals (MOs) and NBOs. When iodine acts as an acceptor, bonding collinearly in the axial position of a square-planar d(8) Pt(II) complex, the dominant contributor to the bonding is a sigma*(I-I) orbital as the acceptor orbital, while a mainly d(z)(2) orbital centered on the metal center is the corresponding donor. That this kind of bonding is characteristic of axial bonding in d(8) complexes was supported by model calculations with incoming donors and acceptors, NH3 and BH3. In contrast, the distinct "bent" coordination of the I-2 bound at the axial position of the [Rh-2(O2CCF3)(4)] paddle-wheel complex is associated with a dominant donation from a p-type lone pair localized on one of two iodine atoms, the sigma*(Rh-Rh) antibonding orbital of the metal complex acting as an acceptor orbital. We check the donor capabilities of I-2 in some hypothetical complexes with Lewis acids, H+, AlCl3, B(CF3)(3). Also, we look at the weakly bound donor acceptor couple [(I-2)center dot(I-2)]. We explore the reasons for the paucity of I-2 complexes and propose candidates for synthesis.